Professional Documents
Culture Documents
2-1
Regardless of the process by which the existing allocation of water in a basin was established, it
may be helpful to compare that allocation with an efficient allocation because: 1) such a
comparison may indicate promising opportunities for private water trades; 2) where trades are
hampered or precluded by institutional barriers, the comparison may indicate areas where
institutional reforms may allow for a more efficient water allocation; and 3) where past water
developments were publicly financed, the comparison may indicate changes that the public
entity should consider to increase the efficiency of the public project. Also, when considering a
new publicly-financed water development, the opportunities for the project to increase the
efficiency of water allocation should be carefully analyzed.
Economic Efficiency
Two fundamental economic concepts are efficiency and equity. Economic efficiency is
concerned with the aggregate wealth generated in a society; economic efficiency is improved if
aggregate wealth increases. Equity is concerned with the fairness of the distribution of that
wealth; equity is improved if fairness increases. Economic efficiency is subject to technical
evaluation, given the assumptions of the dominant (neoclassical) economic model. Equity is not
subject to technical evaluation within the neoclassical model. Therefore, regarding equity, the
role of economics within this theoretical model is to indicate the effect of a policy on income
distribution so that equity can be considered in the political arena.
Three assumptions of the neoclassical economic model are particularly relevant. The first is that,
for the purpose of evaluating economic efficiency, the existing income distribution is accepted as
given. This is important if a different and more acceptable income distribution would result in
different demand and supply schedules, and thus in different prices for goods and services.
Acceptance of the existing income distribution is justifiable in a democracy because the income
distribution is, to a large extent, a matter of public policy.
The second assumption is that consumers are the best judges of their welfare; thus, the values
that individuals assign to goods and services are unquestioned. This consumer sovereignty
assumption is not fully met, if for no other reason than that consumers often lack accurate
information. However, in a democratic society, consumer sovereignty is not an unreasonable
premise upon which to base resource allocation. Establishment of a democracy is, in a sense, a
repudiation of allowing a select few to control resource flows.
The third assumption is that impacts outside the boundary of the analysis can be ignored. In water
resource literature, the boundary of the analysis has been called the "accounting stance." The
accounting stance separates the geographic or demographic area of concern from that which is
ignored. This choice of boundary is important because it can determine whether a particular benefit
or cost is relevant to the analysis. For example, effluent that lowers downstream water quality is
ignored if the downstream area is beyond the boundary of the analysis. Typically, the accounting
stance coincides with the administrative responsibility of the entity performing the analysis, but
might be restricted by such things as data availability or the capacity of an analysis tool.
2-2
If benefits and costs occur at different points in time and the timing of their occurrence matters,
the efficiency criterion must account for time. The traditional approach to this problem is to
compute a present value (PV). An allocation of resources across the total number of time periods
(np) is efficient if it maximizes the PV of net benefits that could be received from all the possible
ways of allocating those resources over the np periods. Mathematically, this is accomplished by
evaluating the discounted algebraic sum of a stream of benefits minus costs over the life of the
project or period of analysis, as follows
(2.1)
where Bi is the benefit and Ci is the cost in period i, np indicates the total period of analysis, and
r is the discount rate. Choice of the discount rate may depend on the opportunity cost of capital
and can include a risk adjustment. Typically r is below 0.1. If the future will not be discounted, r
is set to zero. The current version of Aquarius does not compute discounted benefits, although
provisions have been made in the model for its future implementation. In what follows, we
assume a zero discount rate.
It is important to recognize the distinction between economic and financial efficiency. Financial
efficiency is concerned only with money flows. From this standpoint, a good or service is of
value only if money changes hands when it is exchanged or consumed. Economic efficiency is
concerned with all goods and services valued by the public regardless of whether consumption is
accompanied by monetary exchange. Financial efficiency is improved if net financial return
increases. Economic efficiency is improved if the net wealth of society increases. For example,
additional instream recreation opportunities available without cost to users are ignored from the
standpoint of financial efficiency but are relevant to economic efficiency if recreationists would
be willing to pay for the opportunities if they had to pay to use them. Financial efficiency is a
primary concern of private firms and corporations but may also be of interest to governments
evaluating their cash flow for budgetary purposes. Economic efficiency is typically a primary
concern of governments or organizations charged with administration of public welfare.
Aquarius can be used to evaluate either financial or economic efficiency. A key difference
between evaluating financial and economic efficiency involves the selection of water uses to be
included in the analysis. Financial efficiency may ignore some uses that are relevant to economic
efficiency. For example, an individual firm may not care about the downstream effects of its
water depletions, but from the economic efficiency standpoint those depletions are important
considerations as long as the downstream area is within the spatial purview of the study.
Economic efficiency is maximized when the sum of benefits minus costs from the various
outputs is as high as possible given people's valuesas expressed by their willingness to pay or
willingness to accept compensationand the physical constraints of the system (the production
functions). In the remainder of this chapter, we assume that costs are nil and that water uses are
independent (i.e., willingness to pay for a particular water use is independent of willingness to
pay for other uses). We also assume a Marshallian demand function rather than the more precise
Hicksian (income compensated) demand function. A Marshallian demand function will closely
approximate its Hicksian equivalent if income effects are small (Willig 1976). Income effects
2-3
should be small where the good at issue is a very small portion of the total purchases of the
individual, as is generally the case with water.
Given these simplifications, P is price (i.e., marginal willingness to pay) and x is level of output
in the following demand (i.e., marginal benefit) function: P = f (x). With demand functions for
the various water uses (j), the total benefit function (TB) to be maximized over the various time
periods (i) is
(2.2)
where np = total number of time periods, nu = total number of water uses, and a = level of consumption.
TB is maximized when ai j are set so that the Ps are equal for all i and j. In other words, total
benefits are maximized when the levels of consumption are such that the marginal benefits for
each use across all uses and time periods are equal. TB can only be maximized over the j uses for
which marginal benefit functions are specified. If relevant uses are omitted because their benefit
functions cannot be specified, they must be represented in some other way such as by adding a
physical constraint to the model.
2-4
It is common to describe a point along a demand function in terms of its price elasticity of
demand (hereafter just elasticity or ,), which expresses the percentage change in quantity
demanded for a percentage change in price. At a point along the curve, , = (MQ/Q) / (MP/P)
where Q is quantity demanded and P is price. For a downward sloping curve, an , < 1 is
elastic, and an , > 1 is inelastic. The special case of , = 1 is unit elasticity. A linear
downward sloping demand curve is elastic in the upper left and inelastic in the lower right, with
a point of unit elasticity in between. A constant elasticity downward sloping demand curve is
convex to the origin (figure 2.1b), and is undefined at a zero price. Such a curve is represented
by a power function such as Q = c P d with c > 0 and d < 0. The exponent on price d is equal to
,. A demand curve in power form can be estimated by knowing , and one price-quantity point
along the curve. Because the curve does not cross either axis, the ends of the curve may be very
sensitive to the values of the point specified.
An attractive alternative to the power function is the exponential function, which can
approximate a power function and has the additional feature that the curve crosses an axis. The
demand function Q = g eP/h, with g $ 0 and h < 0, is downward sloping and causes the demand
curve to intersect the quantity axis, indicating that satiation is reached. The inverse demand
function P = a e Q/b, with a $ 0 and b < 0, is downward sloping and causes the curve to intersect
the price axis indicating that consumption ceases if the price is high enough (figure 2.1c).
Aquarius uses the inverse demand function.
The exponential function can be estimated by knowing two points along the curve. One point
that may be known, or at least more easily assumed than others, is the price at which quantity
demanded falls to zero. Call this price PQ=0 . Another convenient point is given by the quantity at
which price equals the existing price Pe, call it QP=Pe. Given these two points, the intercept a =
PQ=0 and the exponent b = QP=Pe /[ln(Pe ) ln(PQ=0 )] of the inverse demand function P = a eQ/b.
2-5
The two-point approach to estimating an exponential function fails to use information about
elasticity of demand that may be available for at least one point along the demand curve.
Elasticity varies along an exponential function. Aquarius includes a computational tool to
estimate the coefficients of the exponential function knowing two points and the elasticity at one
of those two points. The tool uses a weighted least-square method to minimize the sum of
squares of deviations of the observed points and elasticity from those of the fitted function. See
Appendix B for details.
The coefficient b of the exponent in the inverse demand function P = a e Q/b is negative if the
curve is downward sloping. However, note that when specifying b in Aquarius the minus sign is
assumed, so that the user enters only the absolute value of b (i.e., P = a e -Q/b) (see Chapter 4 and
Appendix B).
It is important to distinguish between demand for water in general and demand for water from
the particular source being modeled. For example, demand for the first increments of residential
water is extremely high. Thus, a demand curve for residential water in general might intersect
the price axis at a very high price. However, residential water from a particular source may have
alternative, though more costly, sources. The cost of the next more costly source would truncate
the demand curve for the particular source at issue at a price equal to that cost. Whenever an
alternative source exists, its cost can serve as an estimate of PQ=0.
When comparing demand functions for water in different uses it is important that each demand
function be for raw water. For example, it would be incorrect to compare demand for raw water
in irrigation with demand for treated and delivered water in residential use. The latter estimate of
willingness to pay would include the cost of treatment and delivery whereas the former would
not. If the process of estimating a demand function for a particular water use focuses on treated
and delivered water, treatment and delivery costs must be removed from the estimated
willingness to pay to yield demand for raw water.
For offstream water uses that do not completely consume the diverted water, such as most
agricultural and municipal withdrawals, the quantity variable in a water demand function must be
chosen to correspond with the way diversions are being modeled. Sometimes in modeling water
allocation, only the consumptive use amount is modeled; that is, withdrawal is set equal to the
consumptive use and return flow is thereby ignored. This is a viable procedure if actual return flows
reach the stream upstream of the next downstream diversion point of the model. In this case,
consumptive use is the appropriate quantity variable for the water demand function. A more
accurate characterization of water movement in a river basin is achieved by explicitly modeling
return flows and specifying the point where they reach the water course. In this case, withdrawal is
the appropriate quantity variable. Either way, consumptive use includes actual evapotranspiration at
the location of use plus any other losses along the delivery and return paths.
In a well functioning water market, the marginal value of raw water is equal across uses. This
equality occurs because, if marginal values differ, the higher-valued use can afford to purchase
water from a lower-valued use, paying a price that exceeds the water's value in the lower-valued
use. Transfers from lower-valued to higher-valued uses continue until the advantages of trade are
eliminated, that is, until marginal values are equal and an optimal allocation is reached. Equality
of value at the margin does not indicate equality of value over the full extent of the demand
2-6
curve. Indeed, the value of the initial units of water to domestic or industrial uses, for example, is
greater than the value of initial units to most irrigated agriculture. However, in the absence of
institutional barriers to trade or significant transaction costs, actual marginal raw water deliveries
to each water use will occur at the point of equal marginal value.
Markets in the real world rarely perform as smoothly as they do in theory. Water transfers are
often constrained by institutional barriers such as contractual agreements attached to publicly
funded storage and delivery projects that preclude sale to unauthorized uses, or by the lack of
legal recognition of some water uses such as habitat maintenance. Water transfers are also
typically impeded by significant transaction costs such as those associated with quantifying and
legally defending return flow amounts. Institutional barriers and transaction costs may keep
marginal values of water in different uses from reaching equality, requiring separate valuation
efforts to estimate the current marginal values of water in different uses.
In the following sections, we present information that will facilitate use of Aquarius including:
1) evidence on the shape of the annual demand curve; 2) detail on placing the annual demand
curve in realistic price-quantity space; 3) information on the difference between demand for
treated and delivered water and demand for raw water; and 4) information on disaggregating
annual demand to the monthly time step. Prices from studies performed over the past twenty
years or so are reported. We have, unless stated otherwise, adjusted for inflation by updating
these prices to 1995 dollars using the Gross Domestic Product (GDP) implicit price deflator.
Table 2.1 Water use in the United States in 1995 for the major consumptive use types
in millions of gallons per day, with percent of total use in parentheses.a
Western Statesb
United States
Colorado
Withdrawal
Residential
25,888
(8)
7,200
(7)
508
(4)
Commercial
9,512
(3)
3,386
(3)
109
(1)
25,459
(8)
2,723
(2)
143
(1)
Thermoelectric
131,852
(40)
1,160
(1)
128
(1)
Irrigation
133,621
(41)
96,297
(87)
12,735
(93)
Industrial
Consumptive Use
Residential
6,596
(7)
2,147
(4)
154
(3)
Commercial
1,294
(1)
519
(1)
16
(0)
Industrial
3,365
(4)
593
(1)
42
(1)
Thermoelectric
3,313
(3)
317
(1)
41
(1)
81,231
(85)
50,939
(93)
4,915
(95)
Irrigation
a
Source: Solley et al. (1998). Fresh water only, surface plus ground water. Ignores mining and livestock
water use. Percent columns may not add to 100 because of independent rounding.
Includes Arizona, California, Colorado, Idaho, Montana, Nevada, New Mexico, Oregon, Utah,
Washington, and Wyoming.
fields, where other inputssoils, rainfall, management, etc.are roughly the same on the two
sets of fields. This approach yields a point estimate of the value of irrigation water. For an
example, see Duffield et al. (1992, p. 2179). The other variant involves controlled studies at
experimental farms where the effect of water on the production of specific crops is carefully
observed. Two such studies are Ayer and Hoyt (1981) in Arizona and Kelly and Ayer (1982) in
California. These studies estimate production functions expressing the input-output relation
between water and crop production and then infer the value of the water input from the market
value of the crop. Both studies reported that water demand was inelastic for most crops studied
over water prices as high as $100/acre-ft (most elasticities were greater than 0.5). However,
such studies are limited by the combinations of other inputs that the experimental plots were able
to incorporate, allowing for the possibility of more elastic demand in the long-term (i.e., when all
inputs can be adjusted with more flexibility than they were at the experimental plots). This
approach yields demand estimates for individual crops, not for entire multi-crop farms or for
multi-farm regions.
With the programming approach, farmers' production options and input and output prices are
expressed in matrices of linear or quadratic problems. Optimal production decisions are
determined via solution of the corresponding programming problems. This approach is typically
applied to individual representative farms. The results of such runs may be extrapolated to model
efficient allocations across all the farm land in a given agricultural region. Water demand
functions are typically estimated using this approach by solving for irrigation water input under
different assumed water prices (yielding estimates of quantity demanded) or assumed water
availability (yielding estimates of marginal water value). With enough different solutions, a
water demand curve is plotted giving the maximum amount farmers in the region could pay for
increments in irrigation water.
Some studies using the programming approach have presented stepped demand functions
showing the values of discrete quantities of water that would be used by a farm or by all the
farms in the given agricultural region. Each step along the curve represents a different crop or
the same crop in a different situation (e.g., on a different soil). Typically these stepped demand
schedules indicate a nonlinear water demand function, one stepper in the upper left than in the
lower right (Anderson et al. 1973, Bernardo et al. 1987, Bowen and Young 1985, Kelso et al.
1973, and Kulshreshtha and Tewari 1991), but sometimes they show a stepped function that
approximates a linear curve (Gisser 1970, Taylor and Young 1995). Figure 2.2 is an example of
a stepped demand function. This function is more accurately expressed as an exponential
function than as a power function. A regression yields the following inverse demand function:
P = 140 eQ/166,667 (r2 = 0.63). The corresponding power function has an r2 of 0.44.
2-9
Figure 2.2 Derived marginal demand for irrigation water (from Kelso et al. 1973, figure 4-11).
The convex shape (to the origin) of the demand function estimated by many programming
studies reflects the fact that feed grain and forage crops typically occupy most of the irrigated
acreage and yield less profit per acre-foot of water applied than do the more highly valued, small
acreage, fruit and vegetable crops. Thus, the feed grain and forage crops occupy the lower right
portion of the demand curve and the fruit and vegetable crops occupy the upper left portion. The
exact shape of the curve in a given area depends on which crops are planted in that area, which
will in turn depend on physical variables, like the weather and soil, and on prices of inputs and
outputs.
Other programming studies plot smooth curves through the price-quantity points determined in
model solutions, or report equations that represent such points. For example, Heady et al. (1973)
used a large scale interregional model of U.S. agriculture to derive a linear demand curve for
irrigation in the western U.S. The model allowed for substitution of dry land farming for
irrigated farming as the price of irrigation water was increased. Howitt et al. (1980) studied
irrigation in California and produced discrete price-quantity points that very closely traced out a
demand curve convex to the origin. Vaux and Howitt (1984) later used this curve to estimate
demand functions for irrigation in northern and southern California expressed in the form:
Q = j + k P 0.5, where Q = quantity and P = price, with j > 0 and k < 0. In contrast, based on the
work of Moore et al. (1974), Vaux and Howitt (1984) adopted a downward sloping linear
demand curve for irrigation water demand in Californias Imperial Valley. Booker and Young
(1994) adapted earlier programming studies of agriculture in the Upper Colorado River Basin to
express agricultural profit as a second degree polynomial of water input that, when used to
derive water demand, yields a demand function accurately represented by the power function Q
2-10
= c P 0.66, with c > 0. The Vaux and Howitt (1984) and Booker and Young (1994) studies both
showed downward sloping demand functions that were convex to the origin.
The econometric approach has been applied with cross-sectional data representing a large
geographical area over which water costs, perhaps because of pumping depths or application
rates, vary substantially. For example, studies by Moore et al. (1994), Ogg and Gollehon (1989),
and Frank and Beattie (1979) each used data from several western U.S. states. Ogg and Gollehon
regressed water use on water price variables using three different functional forms. The most
successful model used a power function, yielding an irrigation water demand function that is
convex to the origin, with a constant elasticity of 0.26. Frank and Beattie assumed a power
function and regressed the value of agricultural production on a set of explanatory variables.
They found elasticities of demand for irrigation water from 1.0 to 1.5 depending on the
geographical region. Using data from seven Texas counties, Nieswiadomy (1985) regressed
groundwater pumping on explanatory variables including pumping cost and compared linear and
log-log (i.e., power) functional forms. The functional forms differed little in their ability to
explain variance in the pumping rate. The log-log coefficient on pump cost, equivalent to the
elasticity of demand, was 0.8.
All past studies of demand for irrigation water found a downward sloping demand curve. Nearly
all programming studies found a nonlinear demand curve, one convex to the origin. These curves
were generally inelastic, with , generally greater than 0.3, at low water prices and elastic at
high water prices (table 2.2), and can be approximated by an exponential function.
Table 2.2 Characteristics of estimated multi-crop water demand functions for the Western
United States.
Elasticity a
Author
Location
Method
Minb
Maxb
Utah
Programming
-0.01
-7.0
Washington
Programming
-0.14
-0.4
Colorado
Programming
-0.6
-0.6
Western U.S.
Econometric
-1.1
-1.1
Gisser (1970)
New Mexico
Programming
-0.17
-3.0
Western U.S.
Programming
-0.36
-1.5
California
Programming
-0.46
-1.5
Arizona
Programming
-0.09
-1.2
Saskatchewan
Programming
-0.05
-3.1
Western U.S.
Econometric
-0.26
-0.26
California
Programming
-0.1
-10.
Elasticities are arc or point and may be as reported in the original paper or computed from information
presented in the paper. Entries are approximate and may represent a compromise between several estimates
presented in the original paper.
The minimum and maximum values of irrigation water reported for the estimated demand function.
2-11
Of the studies that fitted smooth curves to these results, Vaux and Howitt (1984) used nonlinear
and linear functions and Booker and Young (1994) presented a profit function that yielded a demand
function that we most accurately expressed in an equation in power form. The econometric studies
generally either assume a power functiona common practice in economic studies partly because
the exponent on price in a demand function is equal to the price elasticityor find upon comparison
that such a function performs better in terms of explained variance.
2-12
Because there is considerable variety across the West, some knowledge of local conditions is
needed to estimate PQ=0, Pe and QP=Pe. But to provide an example, assume an exponential
functional form, where PQ=0 = $150, Pe = $15, an irrigated area of 10,000 acres, and existing
annual withdrawals of 3 acre-ft/acre. Using these assumptions, the annual inverse demand
function is P = 150 eQ/13,000.
2-13
a
b
2-14
October
November
December
weather and on whether double cropping is practiced, which also depends on the weather. As
table 2.3 indicates, irrigated farming occurs year-round in the Lower Colorado Basin, which
includes the desert areas of Arizona and California but not the cooler Upper Basin.
A limitation of the proportional approach is that it holds the maximum price, which is the
intersection at the price axis, constant for all months at the level set by the annual demand curve,
when in fact the farmer might be willing to pay more or less in some months. For example,
because in a given month flexibility is limited to substitute among inputs, a farmer might be
willing to pay more for initial water quantities than the annual demand curve indicates. Another
limitation is that the assumption of constant monthly proportions of annual water use precludes
the possibility of adjusting the use among months in response to existing moisture conditions.
Nevertheless, given the lack of estimated monthly demand functions and of data on how
irrigation responds to natural moisture changes, the constant proportion disaggregation approach
is probably the most viable.
Location
Elasticity
Billings (1990)
Tucson, AZ
-0.57
Griffin (1990)
Texas
-0.37
Malmo, Sweden
-0.15
Denver, CO
-0.49
Nieswiadomy (1992)
Northeastern U.S.
-0.28
Southern U.S.
-0.60
Western U.S.
-0.45
-0.14 to -0.44
Rizaiza (1991)
-0.40, -0.78
Columbus, OH
-0.26 to -0.50
Massachusetts
-0.41 to -0.69
United States
-0.49
In most cases, the elasticities were estimated from a constant-elasticity functional form. Where a
variable-elasticity functional form was used, we report the arc or point elasticity that was reported
in the original, which is estimated at the average price for the sample. A range is also listed if
different estimation procedures or different conditions resulted in more than one estimate. Where
both long- and short-run elasticities were reported, only the long-run elasticities are listed. Griffin
(1990), Hanke and de Mare (1984), and the -0.5 estimate of Schneider et al. (1990) were listed as
"municipal;" all others were listed as "residential."
Author
Location
Winter
Summer
Danielson (1979)
Raleigh, NC
-0.31
-1.38
Griffin (1990)
Texas
-0.32
-0.40
Texas
-0.16
-0.38
Grima (1973)
Toronto, ON
-0.75
-1.07
Howe (1982)
-0.06
-0.43 to -0.57
Lyman (1992)
Moscow, ID
-0.65
-3.33
Renzetti (1992)
Vancouver, BC
-0.01
-0.65
Griffin (1990) and Griffin and Chang (1990) were listed as "municipal" and all others were listed
as "residential."
2-16
The estimates of elasticity among the 39 studies include a mixture from short-run to long-run
estimates. Although most studies of water demand elasticity did not attempt to measure both
short- and long-run elasticities, a few studies (e.g., Schneider et al. 1991, Lyman 1992) estimated
both and found that long-run demand is more elastic than short-run demand. This difference in
elasticity reflects adaptations to changing prices that occur gradually as consumers change habits
and alter water using appliances or landscape vegetation.
Nearly all studies of residential or municipal water demand use an econometric approach, most
with cross-sectional data but some with time series data. Most studies assume a non-linear
functional form, thereby predetermining the shape of the annual demand curve. Many studies
assume a Cobb-Douglas (power) function that yields a constant elasticity curve; note the number
of studies listed in table 2.4 with only one elasticity estimate, which indicates a constant
elasticity model. Some other studies (e.g., Foster and Beattie 1979) assume an exponential
function on water price.
Studies of water demand in the U.S., whether cross-sectional or time-series studies, have used
data with relatively little variation in water price. For example, average prices charged by the
218 cities studied by Foster and Beattie (1979) varied across cities from $0.10 to $1.79/m3, and
average prices charged by the 30 Texas communities studied by Griffin and Chang (1990) varied
across communities from $0.34 to $1.62/m3. U.S. studies do not include cases with sufficiently
high water prices to allow accurate estimates of the shape of the upper left portion of the demand
curve. Thus, the estimates of price elasticity reported in the 39 studies apply to the low-price
portion of the residential demand curve.
Martin and Thomas (1986) provided a unique estimate of the entire demand curve for residential
water via their comparison of water use in five cities that, although sharing an arid climate, differ
markedly in the cost of providing water and in the price charged to consumers. These prices varied
from $0.16 to $25.01/m3 and represent cities in the U.S., Kuwait, and Australia (table 2.6).
Table 2.6 Residential water price and quantity data for selected cities.
Quantity
(l/c/d)a
Price
($/m3)b
50
25.01
184
1.19
Perth, Australia
288
0.55
Tucson, AZ
371
0.48
Phoenix, AZ
595
0.16
35
8.39
Author
Location
Rizaiza (1991)e
2-17
Martin and Thomas reported that a demand curve with constant elasticity of 0.49 provided the
best fit of the five points. We found one other high-priced case in the literature (Taif, Saudi
Arabia) reported by Rizaiza (1991) (table 2.6).
Location
Type of
water use
Elasticity
Commercial
Lynne et al. (1978)
Miami, FL
Commercial
-0.11 to -1.33
Columbus, OH
Commercial
-0.92
Government
-0.78
School
-0.96
Commercial
-0.36
United States
Industrial
De Rooy (1974)
New Jersey
Industrial
-0.35 to -0.89
Elliott (1973)
29 US cities
Industrial
-0.64
United States
Industrial
-0.33 to -0.80
Renzetti (1992a)
Vancouver, BC
Industrial
-1.91
Renzetti (1992b)
Canada
Industrial
-0.15 to -0.59
Renzetti (1993)
Canada
Industrial
-0.66 to -2.17
Columbus, OH
Industrial
-0.44
Turnovsky (1969)
16 towns in MA
Industrial
-0.50
United States
Industrial
-0.74
2-19
a
b
April
7.9
6.3
May
8.5
6.4
June
9.5
9.8
July
11.2
14.8
August
11.2
16.8
September
9.9
11.4
October
7.5
7.2
November
6.8
5.6
December
6.8
5.2
demand based on the typical monthly proportions of annual water use. Because indoor water use
is the most valuable portion of domestic use and does not vary with weather conditions, the point
where the monthly demand curve intersects the vertical axis should remain constant across
months. The greater elasticity of demand (table 2.5) and greater quantity demanded (table 2.9) in
the summer months lead to a summer demand curve that stretches further to the right and is
flatter at a given price than one of the winter months.
Hydropower
As Young and Gray (1972) and Gibbons (1986) explain, the value of hydroelectric energy is
commonly estimated using the alternative cost technique, based on the reasoning that electricity
not produced at hydroelectric dams will be produced using the next more expensive process.
Use of the alternative cost valuation method is appropriate because the energy produced at a
hydroelectric plant typically enters a large electric grid, composed of many hydro and thermal
electric plants, that supplies numerous demand areas. The various plants in the grid are
substitute suppliers. If one supplier reduces production, other suppliers make up the difference,
and a small shift from hydro to a more expensive energy source will have a negligible effect on
overall energy prices. Typically, baseload power produced at a hydroelectric plant is replaced by
a coal-fired plant and peaking power is replaced by a gas- or oil-fired turbine plant. In the
western U.S., gas-fired turbines are most common. More expensive oil-fired turbines also play
an important role in the eastern U.S.
Using the alternative cost method, the value of hydropower in dollars per acre-foot is estimated
to be equal to 0.87 feet of head cost savings, where the cost savings is the cost per kilowatt
hour (kWh) at thermal plants minus the cost per kilowatt hour at hydro plants. The 0.87 reflects
the efficiency of converting the energy of falling water into electrical energy. The cost savings
may be determined for the short run or long run. In t mhe short run, capital costs are fixed and
the only costs of note are those for operation and maintenance (O&M). At thermal plants, O&M
costs are dominated by the cost of fuel. Variable costs tend to be constant per unit of output,
allowing simple average costs to be used regardless of the quantity of electricity. Gibbons (1986)
reported that O&M costs at U.S. thermal plants averaged 18.52 and 44.01 mills/kWh at coalfired and gas-fired plants, respectively, whereas the O&M costs at private hydroelectric plants
averaged 1.52 mills/kWh (1980 dollars). The short-run values of hydropower thus averaged
$0.0148 and $0.0370/acre-ft per foot of head for baseload and peaking power, respectively.
Fuel prices are changeable. At about the time Gibbons (1986) published her report, the prices per
kWh that electricity plants paid for fuel dropped, nationally to about 14 mills for coal and 23
mills for natural gas (Energy Information Administration 1996). These lower prices by and large
were maintained for more than a decade. However, natural gas prices have recently (2004-2005)
been rising dramatically; nationally, prices per kWh rose to over 60 mills for natural gas,
although coal prices changed little. Price changes have corresponding effects on total O&M costs
at thermal plants. Accurate estimates of the value of hydropower require careful examination of
costs at the affected thermal plants, along with decisions about how to represent the long-run
trend in fuel prices given data on actual prices that may show considerable short-term variations.
A long-run value would take into account the full costs of both types of plants and would include
capital costs in addition to O&M costs. Estimating these costs is an involved process. Young and
Gray (1972) provided one attempt to estimate the long-run value of hydropower. As Gibbons
(1986) indicated, in an update of Young and Gray's work, the long-run cost savings per kilowatt
hour may be smaller than the short-run cost savings. Most economic studies use short-run values.
2-22
Gibbons (1986) reminds us that using the alternative cost method described above ignores some
costs, such as the environmental costs of thermal plants (e.g., air pollution) and hydro plants
(e.g., inundation of a riparian area), and ignores the differences in the reliability of service
between the two kinds of generating plants. However, similar costs of other water uses (e.g.,
nonpoint source pollution in irrigated agriculture) are also typically ignored when estimating
water value. Such costs should be separately accounted for if economic efficiency is at issue.
2-23
Lagged effects also occur for various recreation activities. For example, flows affect the ability
of fish to propagate (Cheslak and Jacobson 1990), which eventually affects angler catch rates.
Also, flows affect streamside vegetation, which over time affects the quality of streamside
camping or picnicking (Shelby et al. 1992b).
Immediate and lagged effects on the quality of a recreation experience may translate into
changes in individual willingness to pay for a recreation trip (quality effect) and in the number of
trips taken (participation effect). Thus, the recreational value of a given flow level, V(Q), has the
following four effects
(2.3)
Adapting the model of Duffield et al. (1992), the aggregate present value of a change in flow in
the current time period t is given by the partial total derivative of PV with respect to Qt
(2.4)
Table 2.10 Early studies of the economic value of streamflow for recreation.
Measured effects of flow on valueb
Immediate
Author
Bishop et al. (1987)
Duffield et al. (1992)
River
Methoda
Activity
Colorado (AZ)
CV
Fishing; boating
CV
Poudre (CO)
CV
Qual.
Part.
Lagged
Cong.
Qual.
Trip
Part.
X
d
Fishing; shoreline
Fishing+shoreline
Fishinge
Boating; shoreline
CR
Fishing
Taylor (CO)
CV
Fishing
CV
Fishing
Narayanan (1986)
Blacksmith (UT)
TC+CB
Fishing+shoreline
Seven in CO
CV
Fishing; boating
Ward (1987)
Chama (NM)
TC
Fishing+boating
Combined current and lagged effects without accounting for interactions between them.
2-25
CV = contingent valuation method; TC = travel cost method; CR = cross-sectional analysis; CB = contingent behavior (i.e., contingent visitation).
Qual. = quality effect; Part. = participation effect; Cong. = congestion effect.
c
Dollars per acre-foot at selected flow levels. The relation usually applies across months of the season; Daubert and Young is an exception
as they estimated separate relations for each month of the fishing season.
d
Picnicking, camping, hiking, and relaxing.
e
Many in U.S.
Flow
volumec
Fish
caught
X
X
X
X
X
X
Two studies (Daubert and Young 1981, Hansen and Hallam 1991) achieved a value estimate that
to some extent included both immediate and lagged effects. However, even in these cases
interactions between immediate and lagged effects, such as the effect of current fishing pressure
on future fish populations and catch rates, were ignored.
Most of the studies in table 2.10 calculated a recreational value of instream flow, usually in
terms of acre-feet. Five of the studies estimated the change in marginal value of flow as flow
changed, but others only reported average values or marginal values for a common flow level.
All else equal, unit value of flow is expected to be greater the: 1) more recreation activities
occur, 2) longer the stretch of river that receives recreation, 3) greater the number of trips
received per activity per mile of river, 4) better the quality of the recreation experience, and 5)
less is the river flow level. Bishop et al. (1987) reported an average value of about 50/acre-ft for
fishing and boating recreation downstream of Glen Canyon Dam in the relatively high-volume
Colorado River. Hansen and Hallam (1991) reported a wide variety of current marginal values
across rivers, but values above $30/acre-ft were uncommon.
2-26
Nonuse Values
Nonmarket, or noncommodity, uses of water include not only recreation (discussed in the
previous section) but also preservation. Whereas riparian recreation involves an actual trip to a
river or stream, preservation of riparian areas and aquatic habitat does not. Nevertheless,
preservation is valuable to many people, as evidenced by the millions of dollars annually
donated to organizations such as the Nature Conservancy. Preservation is known by economists
as a nonuse value.
Using contingent valuation, economists have attempted to estimate preservation value, focusing
on categories of nonuse value called existence and bequest values (see Brown 1993 for a survey
of nonuse value studies). A few of the nonuse value studies have focused on water flow. For
example, Loomis (1987) studied the nonuse value of flows into Mono Lake, and Brown and
Duffield (1995) studied the nonuse value of preserving flows in two Montana rivers. However,
application of contingent valuation to estimate nonuse value remains controversial (Arrow et al.
1993, Portney 1994). No doubt nonuse values for preserving certain species or ecosystems are
substantial; however, there is no consensus that current methods accurately measure such values.
In the absence of nonuse values for preservation of streamflows, preservation concerns can be
handled in Aquarius using physical constraints.
2-27
Final Remarks
Specifying demand curves is a difficult task. Although we know much about demand for different
water uses, the precise location of a given demand curve is difficult to estimate (as evidence,
observe the variety of elasticity estimates in tables 2.2, 2.4, 2.5, and 2.8). All attempts to plot
demand curves must be regarded as approximate. In many situations, much extrapolation of
demand in other locations, where demand studies have been done, is needed in order to proceed.
Nevertheless, price-quantity relations do existthough they may be difficult to precisely
plotand they do affect actual water allocation. Our understanding of efficient water allocation
will be enhanced to the extent that we can model water allocation based on reasonably accurate
demand functions. Even suboptimal solutions (based on approximate demand functions) are
useful for understanding water allocation if they include all relevant water uses. With the growing
demand for riparian recreation, this nontraditional water use can be equated with traditional uses
in terms of how it is represented in the model and how an efficient allocation is reached.
In this chapter we have provided information about the shape of typical demand functions for
different water uses to facilitate the task of specifying demand curves when good site specific
demand studies are lacking. Because any such specification is an approximation, sensitivity
analysis is essential if the resulting optimization is sensitive to the demand specifications.
Aquarius has been designed with such sensitivity analysis in mind.
2-28
Chapter 3
This general definition aptly characterizes a water resource system. The elements interacting within
the system, such as reservoirs, irrigation demand areas, and instream recreation activities, are called
system components. System boundaries are defined by the flow network, which the analyst creates
to represent the river basin to be modeled. The network specifies which components will be part of
the system. Inputs, which consist of information necessary to simulate the operation of a system
include, among many others, natural flows, water demands for irrigation, instream flow
requirements, and the price of energy. As explained latter, some of the inputs can be controlled by
the analyst, but others are portrayed as uncontrollable inputs. Outputs are quantitative measures of
the system performance that consist of many component objectives. Depending on the objectives,
performance may be measured in terms of monetary values, amount of energy produced, biological
indexes, or amount of water supplied to offstream demand areas.
Aquarius is a modeling system that helps create a conceptualization, and usually a simplification,
of a real-world system. The degree of detail with which each component is characterized varies with
the desired accuracy. For a model to reflect the real behavior of a system with accuracy, it must
preserve all the essential operational characteristics of the river basin. For a generalized model, such
as the one proposed here, this is a challenge.
3-1
Water Sources
Two water sources are typically found in a river basin: surface and
subsurface water. Both can be represented by Aquarius. Although they
are parts of the same resource, differences in their availability, cost and
administration often cause them to be treated separately. For many years
hydrogeologists and engineers have been developing water supply
sources for a vast array of purposes. As the demand of water increased,
new water sources had to be developed. Only during the last few years, and as a result of a larger
environmental awareness and an increase in the cost of providing additional water, there has been
a change in paradigm, where the development of new water sources has been subordinated to prior
conservation efforts and more efficient use of the water already available.
Surface water, water which remains on top of the land, refers to the natural flows conducted by the
drainage network of a basin, as measured by a gaging station at a given location of a water course.
Flow records can extend from just a few years to 100 years or more. At least ten years are needed
to have an approximate idea of the water yield of the catchment. Records longer than 50 years are
needed to begin to explain in detail the temporal variability of the flow time series at the site. A
surface-water-source (SWS) in general can be either controlled or uncontrolled flows (regulated or
unregulated), although for our purpose a SWS is always interpreted as uncontrolled (unregulated)
flows.
Subsurface water supplies are contained in the saturated portion of the soil or rock profile. An
underground reservoir is usually part of a regional aquifer that may expand much farther than the
boundaries of the catchment of interest. A ground-water-source (GWS) is normally developed from
wells or springs which are usually located near the demand area. Because ground water has to be
pumped from below the surface, it constitutes a controlled water supply.
Where surface and ground water are hydraulically connected, the use of ground water can alter
surface flows, and vice versa. This interrelationship may generate adverse effects on downstream
third parties. Considering the conjunctive use of surface and ground water would add a new
dimension to the water supply component. Aquarius does not model the hydrologic interaction
between the two water sources, although it can simulate the pumping of ground water (GWP) to the
surface.
Rising population and incomes have led to increasing demands for water in agriculture, industry and
other uses, and thus to increasing pressures on water supplies. Legal, institutional and economic
constraints will largely determine whether surface or ground water supplies are first to be exploited.
In many locations the cost of ground water is higher than the cost of surface water due to pumping
costs. But where surface water is scarce, the cost of ground water may be less than it is for surface
water, making augmentation of the supply using ground water sources a viable economic alternative.
However, where the pumping rate exceeds recharge, water tables drop and pumping costs gradually
increase, perhaps shifting the pressure back to surface water supplies.
3-2
Storage Reservoir
The purpose of a storage-reservoir (SRS) is to transform the random and periodic
nature of flows into a series of releases that more closely correspond with the
seasonal water demands in a river basin. This objective is achieved by regulating the
amount of stored water by passing flows through the reservoir outlet works and
spillway to meet downstream water supply functions and to prevent flood damages.
Flow regulation takes an uncontrolled flow, such as water flowing naturally in an undeveloped
upstream basin, and turns it into a controlled flow, such as releases from a reservoir, to satisfy a
particular demand. Figure 3.1 shows a sketch of a reservoir with all possible inflows and outflows
as considered by Aquarius.
XI
du
= controlled inflows
= upstream decision variable
UI = uncontrolled inflows
NF u = upstream natural flows
L u = upstream reservoir spillage
XR
d
= controlled releases
= decision variable
UR
L
E
= uncontrolled releases
= reservoir spillage
= reservoir evaporation
Figure 3.1 Variables used to describe storage
dynamics in a reservoir.
(3.1)
When equation (3.1) is used to describe the dynamics of storage in a reservoir, dS(t)/dt is the rate
of change of the reservoir storage with time, I(t) is the rate of inflows into the reservoir, and O(t) is
the rate of outflows from the reservoir including controlled releases, spillage, and evaporation and
seepage losses.
Analytical optimization techniques require that flows be classified as either controlled or
uncontrolled flows. In figure 3.1, any upstream controlled flow, du, which reaches the reservoir
under consideration, becomes a controlled inflow to the reservoir. If several d u flows exist, they are
all grouped under the term XI. Spillage from upstream reservoirs Lu and natural flows NF u are
considered uncontrolled inflows to the reservoir and are grouped under the term UI. As with d u, the
inflow variable Lu, and NF u may be indicative of one or more sources of flow. Note that all inflows
with the superscript ( u ) originated upstream from the reservoir under consideration.
3-3
Among reservoir outflows, we present two groups: the controlled releases d and the uncontrolled
reservoir releases, such as spills L and evaporation losses E, grouped under the term UR (not depicted
in figure 3.1). As more than one controlled release d is possible, we group them under the term XR.
Controlled releases encompass water for offstream uses such as hydropower, municipal water supply
and irrigation demand areas. Controlled releases for instream uses include flood control in floodprone river reaches, water-based recreation in rivers and lakes and reservoir releases to protect the
integrity of a river ecosystem.
Using the notation indicated above and substituting the time index i for t to indicate discrete time
intervals, the inflow to a reservoir during any time period i is expressed as a function of
(3.2)
(3.5)
Evaporation rates can be estimated from pan evaporation data available for the region where the
reservoir is located. Because precipitation over the lake is another source of inflow, the analyst
should consider the net-evaporation rate, obtained as the difference between the pan-evaporation rate
minus the expected precipitation depth during the season. The consideration of seepage losses
should be decided in a case-by-case basis. In general, assume that the amount of water lost by
seepage increases with the level of storage in the reservoir because of the increase in the hydraulic
gradient with additional storage. The model does not account explicitly for seepage losses, although
seepage losses can be imbedded into the evaporation losses as a first approximation.
Aquarius considers a single storage pool for
flow regulation. This volume is defined
between the maximum Smax and minimum
Smin allowed operating storages (figure 3.2),
which are prescribed by the user. For ungated
spillways the maximum allowable storage Smax
equals the storage at the overflow crest level
Screst, also prescribed by the user. In addition,
the user must indicate the reservoir storage at
the beginning of the operation and that
required at the end of the optimization
horizon, which are denoted as S o and S f ,
respectively. Reservoir storage can be
converted into reservoir elevation by applying
(3.4a).
The single regulation pool in Aquarius contrasts with some other reservoir simulation models, such
as HEC-5 (HEC 1989), where the storage in each reservoir is discretized into levels for operational
control purposes. In fact, no more than a single operating pool is necessary when an optimization
algorithm is used to determine the system operation, because the model finds the optimal strategies
based on perfect foreknowledge of inflows and demands. Nevertheless, future versions of the model
may define several operational levels in a reservoir, such as those included in HEC-5 (inactive,
buffer, conservation, and flood control pools), to be able to simulate the historical system operation
based on a set of reservoir operation priorities (reservoir operation rule curves).
offstream releases;
a third discharge at the center, for
instream releases.
In Aquarius water demands that are directly linked to a reservoir (i.e., there is a single link
connecting the reservoir to the demand) put a water call on the stored water. On the other hand,
water demands that are not directly linked to a reservoir providing water to the demand (i.e., there
is more than one link between the reservoir and the demand, so they are indirectly linked) do not
automatically allow a call on the stored water. To allow for a call in this case the user must
implement a special model component called an indirect-water-call (IWC). This dummy component
enables the mathematical connection between the reservoir and the indirectly-linked demand. The
demonstration networks in Figure 6.3 and Chapter 8 provide examples of the capabilities of this
singular element.
Aquarius computes the outflowing discharge Qvalve of a reservoir using the hydraulic equation for
flow in a closed pipe system
(3.6)
where A is the area of the conduit, KL is the loss coefficient for the whole discharge system, and HT
is the total head needed to overcome the various head losses to produce discharge. For details on the
hydraulic calculation of (3.6), in particular the term KL, see USBR (1977).
The capacity of the spillway to pass a flood wave during a given time interval depends on the water
surface elevation in the reservoir forebay at the beginning of the interval and on the length of the
simulation time step. Iterative computational procedures are needed to estimate the gate-controlled
spillway discharge coefficient C and to compute the residual reservoir storage volume at the end of
the simulation time step. However, detailed simulation of the spillway operation becomes relevant
only for small-time simulation steps (hourly, daily). For monthly time steps, and even weekly at
times depending on the size of the reservoir and inflow conditions, detailed operation of the spillway
is usually omitted. For monthly time steps it is convenient to pass any excess flood water
downstream of the reservoir as a block, without attempting any spillway operation procedure.
Aquarius computes the total outflowing discharge Qspill of a reservoir through an ungated spillway
by means of the following power function
(3.8)
where the only new variable is He, the total head on the crest. Knowing the water surface elevation
in the reservoir (obtained from the elevation-storage function) and the elevation of the spillway crest
Hcrest (level of zero outflow), the model computes He and the volume spilled during the time interval
of analysis. This volume is checked against the maximum discharge capacity of the spillway to avoid
the overtopping of the dam. The user should prescribe the basic information indicated in (3.8).
3-7
As indicated in the previous subsection, the central outlet of a reservoir may play the dual function
of releasing water to satisfy downstream demands and at the same time releasing excess water
during extremely rich water periods (the reservoir water surface can be at any level above the
minimum operational level Smin). Thus, the central reservoir outlet becomes a mandatory downstream
connection for all reservoirs. The outflow equation is given by (3.6).
In addition to the capability of the reservoir to release excess water through the emergency structures
introduced above, Aquarius includes a conceptual system component, termed the spill-controller
(SPC), which helps minimize reservoir spillages. The spill controller acts upon the reservoir to
which it is connected to discourage spillages, and indirectly, over all the system components
upstream that provide inflows to the reservoir under consideration (more details about the
characteristics of this component are given in Chapter 5). The spill controller does not perform a
detailed simulation of the spillway operation. In fact, operation of a gate-controlled spillway is
unnecessary for monthly time steps of simulation, as in this model (the model passes any excess
flood water downstream of the reservoir as a block).
Hydropower Generation
Operation of an electric power system involves considering three basic components
or subsystems, the: i) energy sources (thermal, hydro, nuclear); ii) electric network
components (generators, transformers, transmission lines); and iii) energy demand
or load, constituted by a large number of electrical devices connected by the
customers to the system. Each one of these components is a highly specialized
engineering field that may require detailed modeling efforts depending upon the problem under
consideration (e.g., El-Hawary and Christensen 1979). Aquarius deals only with the energy source
component of hydropower (HPW).
In hydropower systems, hydraulic turbines convert the energy in water stored in reservoirs or other
structures into kinetic energy, which in turn is converted into electric energy by the generators.
Hydropower facilities are typically hydraulically linked to reservoirs, and it is convenient to model
the power plant and the reservoir as a unit. Aquarius can represent the following types of
hydropower installations:
C storage plants, with considerable storage capacity in the reservoir that feeds the powerplant;
C run-of-river plants, with negligible storage capacity and use water as it becomes available.
A hydropower installation is treated as a variable-head powerplant when the net hydraulic head
acting on the turbines changes with water surface elevation changes in the reservoir. Typical layouts
of variable-head plants show the powerhouse as: an integral part of the dam; a separate unit
constructed at the toe of the dam (figure 3.5, left); or built at a distance from the dam and connected
to the reservoir by tunnel and penstock. A fixed-head powerplant operates under a nearly constant
net hydraulic head and is connected to the reservoir by a canal or some other type of open-channel
structure (figure 3.5, right). This is also called a canal powerplant.
3-8
Figure 3.5 Layouts of storage hydro plants; (left): variable-head, (right): fixed-head.
Flow regulation at the reservoir requires a decision to control the amount of flow being allocated
from the reservoir to the storage powerplant. The model recognizes any link transporting water from
a reservoir to a powerplant as a decision link, and creates the corresponding set of control variables.
The second basic type of power stations, run-of-river plants, are located in or alongside the main
water course. Although several different layouts of run-of-river plants can be encountered, they all
present: a diversion structure across the river with its intake; a canal that conveys the water into the
forebay of the powerplant while gaining hydraulic head; the power-house; and another canal that
routes the flow back to the river. The energy head acting on the hydraulic machinery is typically
assumed fixed. This type of installation has insignificant impoundment at the powerplant. Therefore,
water that cannot be passed through the hydro turbine runners because of a turbine shut down is
spilled.
It might be assumed that because flow
going into a run-of-river plant is
unregulated, the generated electric power
is not a controllable variable. However,
when the same run-of-river plant is
analyzed within the context of a multi-user
water system, it is necessary to make a
decision as to how much water should be
diverted into the powerplant and how much Figure 3.6 Power diversion in conflict with instream use.
should remain in the river to comply with
some instream use. As an example, figure 3.6 shows a fishery located in a river reach between the
points of diversion and return of flows from a powerplant. The water allocation conflict between the
two uses arises because of the need to limit the amount of flow that can be diverted into the
powerplant in order to protect the fishery.
3-9
The electric energy output of a hydropower plant is primarily a function of two variables: the
hydraulic head and the rate of water passing through the turbines. The rate P, in kilowatts (kW), at
which electrical energy can be generated is approximated by
(3.9)
where Q is the rate at which water is discharged by the powerplant in cubic meters per second
(m3/s), Hn is the net (or effective) hydraulic head on the turbines in meters, 0t is the turbine
efficiency, and 0g is the generator efficiency. Both efficiencies can be grouped into an overall
efficiency 0, which depends on discharge and effective head. The net head Hn in (3.9) is equal to
the gross head minus energy losses before entrance to the turbine and outlet losses. The gross head
of a powerplant is defined as the difference between forebay elevation and tailwater elevation.
Customarily, the capability of a powerplant to produce energy is expressed by the energy rate
function (erf), which calculates the amount of energy (in kilowatt hours) generated by the plant per
unit volume of water released through its turbines (in cubic meters) during a unit period of time (one
hour)
(3.10)
As presented later, the modeling of power production requires a continuous and differentiable form
of the erf. As a first approximation, we can use the simple analytical expression given in (3.10),
where the effect of discharge is implicitly considered within the variation of Hn and 0. If necessary,
a family of curves parametric of the discharge Q can be used to represent the surface in figure 3.7.
When backwater effect from downstream reservoirs affects the tail-race elevation, special modeling
considerations should be adopted (not available in the model).
3-10
Agriculture Irrigation
The lack of precipitation in many arid and semi-arid regions precludes crop
production unless moisture is supplemented with irrigation (IRR) water. Historically,
the need to reduce these moisture deficiencies has led to development of agricultural
water supply projects, which are the largest consumptive users of water in a river
basin. In the Western United States, such projects store runoff that occurs during the
spring snow melt and releases it during the drier period of the growing season.
Large irrigation developments are composed of a complex array of water related structures for flow
regulation, diversion, conveyance, distribution, and application. Once the irrigation water is applied
to a field, additional structures may be required for capturing and transporting return flows back to
the main water course. Aquarius deals with the first and last stages of the agricultural water use
system: i) the supply of water to the demand area, and ii) the physical link to return the nonconsumptive portion of that supply to the system. The model does not analyze the intricacies of the
distribution and application subsystems.
Water for irrigation may be diverted directly (i.e., without prior storage) from a river to a farm if
natural streamflow is sufficient. However, low flows in a river may severely limit the capacity of
a natural system to reliably supply water, particularly for large irrigation developments. By creating
a storage capacity between the water source and the demand area, the reliability of the water source
increases dramatically. The net increase in dependable water supply, the firm yield, is a direct
function of reservoir storage. Typically, a storage capacity that supplies irrigation water holds a
portion of the incoming water in storage during the high flow season until the natural flow recedes
to the point where it is no longer sufficient to meet the downstream irrigation demand. Releases of
stored water from the reservoir are then used to augment flows in the river to meet the demand.
Often the need for augmented flows persists until the end of the growing season. Even during the
winter months, small amounts of water may still be allocated to the demand area for domestic use,
stock water, or other purposes.
In addition to surface water, ground water can be used to satisfy the demand of the agricultural
sector. When the marginal cost of acquiring surface water for satisfying irrigation demand increases,
ground water may become an economical alternative. In summary, Aquarius can simulate the
following possible schemes for providing water supply to an agriculture demand area:
C direct water diversion from existing flows in the river to an irrigation area;
C water released from a storage reservoir into a river and then withdrawn at a downstream
C
C
diversion point;
water released from a storage reservoir into a canal or conduit, rather than into the riverbed, and
conveyed directly to the demand area;
ground water pumping to an irrigation area.
Whether irrigation water is pumped from an aquifer, diverted directly from a reservoir, or from a
river channel (using a diversion dam), the model individualizes the decision links in the network that
3-11
are necessary to convey water to the demand areas and automatically creates the necessary sets of
decision variables to control the supply.
The amount of water to be supplied to an agricultural area (water duty) must be specified by the
analyst as Aquarius does not include an agricultural consumptive use submodel to estimate water
demand for a given crop and condition. Annual water duties in the Unites States may range from
about 3,000 to as much as 20,000 m3 /ha (1 to 7 acre-ft per acre). Losses in conveyance structures
from the point of diversion to the point of application should be estimated by the analyst and added
to the water duty. The return flow of water from irrigated lands may be collected in drainage
channels where it flows back into a natural river channel at a known location. Alternatively, return
flows may percolate into the regional aquifer to emerge as base flow at some location downstream
in the river (distributed inflow). This return flow augments the prevailing river flow and, depending
primarily on water-quality parameters, the return flow may be reused downstream by other users.
As reported in the literature, the amount of return flow averages about 50 percent of the water
diverted for irrigation purposes, varying from about 20 to 70 percent.
Return flows in the model are computed as a fixed percent of the flow allocated for the period, a
percent (provided by the user) that is assumed constant for all seasons of the year. The model does
not consider delays in the routing of flow return from the agricultural area to the site of recapture
by the flow network. That is, the return flow is made fully available downstream during the same
time interval of simulation. A more realistic conceptualization of return flow routing, necessary
when considering return flows that move as subsurface flow and short simulation time intervals, will
be included in a future version of the model. Whether return flows are transported back to the main
watercourse using a conveyance structure or as ground water accretion, the location in the flow
network where return flows are regained should be indicated by the user during the process of
creating the flow network. The return link is connected to the natural channel using a junction node
(explained in more detail later). If no link is provided for return flows, applicable when representing
a transbasin diversion, the total amount of water diverted is assumed consumed in the demand area.
Once the area under cultivation is defined, the maximum and minimum amounts of irrigation water
required for consumptive use is fixed. Early estimates of the amount of agricultural water available
for the growing season are used to
predetermine the planting area. Then, the
amount of water required to meet the demand
for growing the crops for the entire season is
also defined. Water should be supplied period
after period during the whole growing season
to cover the expected deficiencies in
consumptive water use. Failure to supply the
demanded water at any period may cause the
complete loss of the area under cultivation,
with corresponding economic loss. Figure 3.8
shows a typical pattern of water requirements
for agricultural water in the Rocky Mountain
Figure 3.8 Seasonality of agricultural water use.
3-12
region of the Western U.S., with the maximum demand occurring at the end of the summer. In
contrast to the well defined growing season shown in figure 3.8, other agricultural areas have a yearround growing season. Aquarius allows the user to prescribe the total amount of water to be
delivered to an irrigation area during the entire growing season, the seasonal pattern of the
deliveries, and even the maximum and minimum volumes at each time interval.
3-13
pipes. The four schemes of water supply listed previously for agricultural water can also be
simulated for M&I use by Aquarius. Whether M&I water is diverted from reservoirs, diversion dams
or directly from a river or aquifer, the model detects decision links in the network that convey water
into M&I demand areas and automatically creates the necessary sets of decision variables to control
the supply.
Return flows from urban areas are collected by a network of sewers and transported to treatment
plants before being finally restored to the river channel. It is generally reported that between 60 and
90 percent of the total water supplied to an M&I area is returned as waste water, with small changes
in water quality after treatment. The return flow augments the prevailing river flow and, depending
on water-quality parameters, it may be reused downstream by some other water users. Return flows
in the model are computed as a percent (entered by the user) of the total allocated inflow; this
percentage is held constant for all seasons of the year.
Losses in conveyance structures from the point of diversion to the point of use should be estimated
by the analyst and added to the water demand. As indicated earlier for irrigation return flows, the
present version of the model cannot delay flows returning from the demand area to some later time
step, although for M&I use this is less critical. While building the flow network, the user should
indicate the junction site where M&I return flows are routed back to the main watercourse (using
a left- or right-bank junction node). If no link for return flows is provided, the model assumes that
the total amount of diverted water is consumed in the demand area.
management has moved from ignoring instream recreation to recognizing the impact of
streamflows for recreation and esthetics purposes.
Because the value of stream-based recreation is related to the level of instream flow,
quantification of the relation of instream flow to the quality of recreation for activities, such as
rafting and kayaking, is becoming an important consideration in operating reservoirs and
scheduling releases on many rivers (see Brown et al. 1991).
The attention given to instream flows goes beyond the demand for stream-based recreation.
Colby (1990) comments that additional benefits are generated by the maintenance of instream
flows, benefits related to the knowledge that the riparian ecosystem is preserved (nonuser values)
and to a stream's waste assimilation capacity (water quality benefits). Such benefits are often
included with recreation benefits as conservation or environmental benefits. Because such
inclusion is confusing, following the recommendation of the National Research Council (1992),
we separated water-based recreation from water use to protect ecosystem effects, fish and
wildlife management, and habitat preservation, which is discussed later. The water uses
discussed in this section are the more commercially oriented water-related recreational activities,
specifically, those occurring on natural waterways, excluding reservoirs and lakes.
The growing number of requests from the local community prompted the Tennessee Valley
Authority (TVA) to re-examine the operation of its large network of river-reservoir units, many
of which were originally single-purpose power projects. The resulting study (TVA 1990) is a
major effort to find a socially acceptable balance between the traditional uses of water in the
TVA system, such as hydroelectric power production, navigation and flood control, and
nontraditional uses such as recreation, water quality, and residential development.
A key decision variable in balancing competing water uses is the timing of reservoir drawdown
(figure 3.10). For lake-based recreation and tourism, lakeside residential development, and some
environmental issues, the longer high lake levels are maintained the better. However, delayed
drawdown could cause water spills at the dams creating a risk of substantial hydropower losses,
lack of flexibility in the operation of the power system, and increased pollution introduced by
alternative energy sources.
Recreational use of streams has increased remarkably where whitewater floating is possible. For
example, on the Cache la Poudre River of Northern Colorado, the number of people participating
in commercial raft trips grew from 125 in 1984 to over 30,000 in 1995. During the same time,
private river use by kayakers, rafters, and canoeists increased from a few hundred to about 9,000
trips per year. Similar increases have occurred on many rivers throughout the United States.
In addition to whitewater boating, many rivers have a large number of anglers and participants in
water-enhanced activities such as camping and sightseeing. All this activity has a significant impact
on local economies, where recreationists purchase goods and services related to their water-based
activities. More difficult to measure but equally important is the value of the recreation to the
participants. The value of stream-based recreation can be estimated in economic terms
3-15
commensurate with the values of more traditional water uses by allowing such values to be
incorporated into the multipurpose modeling process. Because stream-based recreation is not sold
in competitive markets like many other water projects, the recreation value estimation requires
specialized approaches. Nevertheless, this should not discourage attention to instream economic
benefits. Recent studies (e.g., Duffield et al. 1992) have provided strong evidence that instream
benefits can sometimes exceed those generated by offstream uses, and that gains are achievable on
some systems by reallocating water from consumptive to instream uses. Similarly, Colby (1990)
argues that instream values are high enough to compete in the market for water rights with offstream
uses when important recreation sites and wildlife species are involved.
Water recreation has marked seasonal characteristics, similar to agricultural water use. The number
of visitors can vary from very low during cold winter months to the very high during the spring and
summer. For recreation, seasonality is also at the daily level as weekday use is typically much lower
than weekends use. This suggests the need to use very short-time intervals when simulating the
operation of a river system that includes recreation activities (Morel-Seytoux et al. 1995).
The use of water for instream recreation is not exempt from conflicts with other uses. For instance,
maintaining higher instream flows during the weekends increases opportunities for activities such
as kayaking and boating but conflicts with electric power objectives because electricity demand
tends to be lower during the weekends. Extra weekend releases may reduce the capability of the
powerplants to meet energy and power demand during weekdays because of lower reservoir levels.
Helping solve conflicts of this nature is the focus of multipurpose water system analysis.
3-16
endangered fish. These models define the necessary flow conditions to protect and enhance
biological species in a river (Lamb 1995), ideally, by re-establishing the natural variability of the
flow regime lost by human-induced changes.
The two most common structural measures causing disruptions in natural flow regimes are
construction of reservoirs that regulate flow and depletion of instream flows by offstream
withdrawals. Figure 3.11 illustrates the effect of flow regulation in a snow-melt dominated basin.
The graph, two mean annual hydrographs for the Black Canyon of the Gunnison River in Colorado,
shows flows before and after reservoirs were built upstream in the mid 1960s. Note the drastic
change in the pattern of flows caused by the presence of the reservoirs. Regulated flows are
significantly higher outside the snowmelt season, and the natural peak in runoff occurring during
the spring and early summer has been totally eliminated.
8000
After Regulation
7000
Before Regulation
6000
5000
4000
3000
2000
1000
0
0
30
60
90
120
150
180
210
240
270
300
330
360
Figure 3.11 Annual hydrographs, pre- and post-regulation, for the Black Canyon of the Gunnison River.
The effects of flow regime alterations on habitat, and of habitat changes on species populations, are
complex. Stream ecologists are, however, gaining a sufficient understanding of these relations, at
least for some species, to recommend changes in water management to maintain habitat. Analysis
is then needed on the feasibility of the changes, and on the effect of these changes on other water
uses. Sometimes such analyses can be accomplished at a monthly time step, but other analyses
require much shorter time steps. For instance, the operation of a peaking hydropower facility can
result in releases from zero (during off-peak hours) to maximum plant capacity (during on-peak
hours). Although average flow values for the day may be adequate based on biological requirements,
the within-the-day flow oscillations would be unacceptable. In addition to water quantity
considerations, water quality is also critical. The release of cold-water from a reservoir can be
detrimental for tailwater fisheries, creating ecological discontinuities in a colony of microinvertebrates that sustain the food web (see Stanford 1994).
3-18
Although economically significant fisheries are often identified, reliable estimation of nonuse and
preservation values in economic terms is not possible. Therefore, Aquarius does not assume that
direct economic values for fish and wildlife habitat are available. This does not imply that this
particular instream use of water has no economic value. Economic valuation research has confirmed
that the American public attaches significant value to wildlife and fish preservation (Boyle and
Bishop 1987), even though that value cannot be estimated with the same precision as other goods
and services. Although the public's willingness to pay for kilowatt-hours of electricity generated,
tons of corn produced, and whitewater kayaking opportunities can be quantified, quantification of
the economic value for the preservation of endangered species and similar environmental concerns
for direct incorporation in this model is insufficient.
What is proposed here is to address the water allocation problem for non-commensurate values in
a unique manner. For a river site where the level of instream flow has been predetermined, but where
the demand function in economic terms is difficult or impossible to define, the analyst can
experiment with fictitious demand curves until the required water allocation level for that instream
use is satisfied. Indirectly, this approach indicates the societal willingness to pay for incremental
increases of flow for difficult to value water uses. In other words, this approach estimates the
economic subsidy required to sustain the activity in open competition with the other directly valued
uses in the basin. Alternatively, the analyst has the option to impose minimum and maximum
instream flow requirements.
Despite the inherent limitation of the model that simulates monthly time-steps only, the capability
of Aquarius to estimate the quantity and quality of habitat at one or several locations in a basin,
along with the other water uses, should enable decision makers to evaluate the adverse effects of
planned or existing water developments and water transfers and identify opportunities for mitigation,
restoration, or enhancement of affected instream and riparian resources.
Flood Control
Flooding, the overflow of water onto normally dry land areas alongside a river or
stream, can cause severe economic and human losses. Flood control is a common
objective of reservoir management, but one often in conflict with other water uses
such as hydropower generation and instream flow requirements. The most general
operational objective of flood-control for a flood prone area or river reach (FCA) is
to minimize the frequency and the magnitude of flood damages along the flood plain by
implementing a series of structural and non-structural control measures. Flood control options
include flood abatement using upstream reservoirs, as well as channel management (dyking,
channelizing) or land use controls (e.g., zoning) in areas subject to flooding (Leopold and Maddock,
1954). Modeling flood control impacts in a river network can help one choose the most effective
combination among those remediation alternatives.
Reservoirs, like natural lakes, can alter the frequency and magnitude of high and low flows (see
hydrographs pre- and post-regulation in Figure 3.11). The ability of a reservoir to suppress flood
3-19
peaks depends largely on its storage capacity relative to the size of the catchment area, and on the
reservoirs operating rules. Flood control reservoirs with planned releases will diminish the
frequency of downstream overbank events. Often flooding risk can be further reduced by operating
a group of storage reservoirs as a system. However, reservoirs will rarely eliminate the risk of
damage from very low frequency high flow events.
Unfortunately, the reduction in the frequency of overbank flooding resulting from reservoir
management of flows causes a deterioration of the channels natural capacity to handle flows, as the
absence of high peaks and flush floods allows sandbar deposition and vegetation encroachment into
the channel. The most common adjustment of the river to the new set of discharges and sediment
load reductions caused by the upstream reservoir is the formation of channel-in-channel sections
with smaller bankfull discharge. This condition in turn exacerbates the flooding problem due to a
decrease in channel capacity to convey high flows.
Nevertheless, protection from floods is a valuable service in many basins, and one that is likely to
grow in importance due to population growth (there is continuous pressure for the reclamation of
flood plains for urbanization) and the annexation of new agriculture land along the river banks.
Moreover, as flow regulation lowers flood peaks in a river reach a false sense of security is created
among inhabitants downstream from the flood control structures, who become unfamiliar with
flooding and expect to be fully protected from all extreme events, further increasing the demand for
flood control.
Access to the benefits of structural flood control measures cannot be restricted to specific users.
Rather, once the structure is provided, the benefits are available to all individuals in the affected
flood plains. Also because access cannot be restricted, payment for the structures must be somehow
shared by all beneficiaries. Typically, flood control structures are planned by government entities
who use tax revenues or property assessments to finance construction and operation of the structures.
Because of the public and nonexclusive nature of flood control, it is usually wise to analyze flood
control impacts and remedies over the entire basin.
Customarily, flood control benefits are estimated as the difference between expected flood damages
before and after mitigation measures are taken. James and Lee (1971, p.251) detail the procedure
to estimate flood damages based on the severity of the floods. Direct damages are losses caused by
physical contact with the flood and are generally evaluated as the cost of replacing or rehabilitating
the affected property in the case of urban damages (private and public) and as the net effect on farm
income in the case of crop damages. The distribution of flood damage potential within the flood
plain may be summarized by a series of land-use maps each indicating the location of various
development classes (agricultural, residential, etc). A second series of maps indicates the severity
of flooding throughout the flood plain for each flow level (flood stage). By superimposing each of
the flood-severity maps over the land-use maps, flood damage as a function of water stage can be
determined throughout the flood plain.
Marginal values for the economic assessment of flood control projects are usually expressed as
functions of the frequency of the design flood. Rather than calculating the mathematical expectation
3-20
of annual flood losses, the continuous simulation of the operation of a river system conducted by
Aquarius provides the opportunity to make explicit the externalities that reservoir discharges impose
on the rest of the system, particularly in flood prone areas. Aquarius can analyze the whole spectrum
of floods to which a basin is subject, providing a more dynamic analysis of the problem.
Water Conveyance
A river reach is the portion of a natural river system conveying flows
between system components. The river reach (RVR), together with
human-made conveyance structures, such as canals and conduits (C&C),
constitute the water system links of a river-flow network.
River reaches can experience different types of open channel flow,
ranging from uniform to nonuniform flow and from steady to unsteady flow conditions, depending
on whether the main flow characteristics are unchanged in space and time. The flow condition
created in a natural channel when a reservoir discharges at varying flow rates creates unsteady flow
conditions in the receiving channel requiring the consideration of time as an additional variable to
describe flow in the river reach. This situation is illustrated in figure 3.12 in which releases from the
upstream reservoir are modified by the river reach before entering the second reservoir.
The movement of a flood wave down a
channel is associated with changes in
timing and attenuation of the wave. The
numerical analysis of this process is
accomplished using some form of channel
routing techniques. Any hydrologic
routing technique is advised for this
model. Hydrologic routing involves
balancing inflow, outflow, and storage
volume through use of the continuity
equation (3.1). A storage-discharge
relation is also required between outflow
rate and storage in the system. There is a
vast body of literature describing
applications of hydrologic routing
techniques (see Bedient and Huber 1988).
When equation (3.1) is used to describe the dynamics of storage in a river channel, dS(t)/dt is the rate
of change of storage within the reach, I(t) is the inflow rate to the reach, and O(t) is the outflow rate
from the reach (figure 3.12). One of the most applied hydrologic routing procedures is the
Muskingum Method, which assumes that outflow from a routing reach is a linear function of the sum
of prism and wedge storage in the reach. Under this assumption, the Muskingum routing equation
yields
3-21
(3.11)
Equation (3.11) indicates that the channel output at a given time step is a linear combination of the
inflow to the reach during the same time step and the inflow and outflow during the previous time
step. The coefficients C1 , C2 , and C3 can be computed once the storage time constant for the reach
K and the weighting factor x, the two Muskingum parameters, are known for the reach under
consideration.
The time required for a flood wave to traverse the reach dictates the need for routing procedures.
When the water allocation problem is simulated using short time intervals of analysis (e.g., 24 hours
or less) streamflow routing is generally required to predict the temporal variations of a flood wave
as it travels from the beginning to the end of the river reach. In contrast, when the time period of
analysis is long (e.g., weekly or monthly periods) it is assumed that all water entering a river reach
is available at the downstream end of the reach within the same time interval of simulation. Hence,
channel routing procedures are unnecessary. This can be verified by knowing the length of the river
reach and a flow velocity for any representative flow rate.
3-22
Aquarius considers any diversion node as a location in the flow network where a water allocation
decision should be made. A diversion link branching out of a node (see again at Figure 3.6) becomes
a decision link when formulating the water allocation problem. In other words, the model converts
the diversion link conveying water to a powerplant, an irrigation area or an urban water demand
zone into a decision link and creates the corresponding set of decision variables.
3-23